Professional Documents
Culture Documents
Pages 599–618
Joseph T. Belter, MS, BS;1* Jacob L. Segil;2 Aaron M. Dollar, PhD, SM, BS;1 Richard F. Weir, PhD3
1Department of Mechanical Engineering and Materials Science, Yale University, New Haven, CT; 2Department of
Mechanical Engineering, University of Colorado at Boulder, Boulder, CO; 3Biomechatronics Development Labora-
tory, Department of Veterans Affairs (VA) Eastern Colorado Healthcare System, Denver VA Medical Center, Denver,
CO; and Department of Bioengineering, College of Engineering and Applied Science, University of Colorado Denver,
Denver, CO
Abstract—In this article, we set forth a detailed analysis of the combination of high functionality, durability, adequate
mechanical characteristics of anthropomorphic prosthetic cosmetic appearance, and affordability. We believe that,
hands. We report on an empirical study concerning the perfor- in order to close the gap, a better understanding of the
mance of several commercially available myoelectric pros- current performance capabilities and performance needs
thetic hands, including the Vincent, iLimb, iLimb Pulse, of anthropomorphic prostheses must be achieved and
Bebionic, Bebionic v2, and Michelangelo hands. We investi-
commonly accepted measures and evaluation protocols
gated the finger design and kinematics, mechanical joint cou-
pling, and actuation methods of these commercial prosthetic
must be established.
hands. The empirical findings are supplemented with a compi- Previous review articles on prosthetic hands have
lation of published data on both commercial and prototype been published [1–4]. Weir provides a thorough discus-
research prosthetic hands. We discuss numerous mechanical sion of prosthesis design, particularly as it relates to chal-
design parameters by referencing examples in the literature. lenges facing people with amputation and their needs
Crucial design trade-offs are highlighted, including number of
actuators and hand complexity, hand weight, and grasp force.
*A portion of this article was published as Belter JT, Dollar AM. Per-
Finally, we offer a set of rules of thumb regarding the mechani-
cal design of anthropomorphic prosthetic hands. formance characteristics of anthropomorphic prosthetic hands. Pro-
ceedings of the IEEE International Conference on Rehabilitation
Robotics; 2011 Jun 29–Jul 1; Zurich, Switzerland. p. 921–27.
599
600
Table 1.
Published general characteristics of commercial prosthetic hands.
Number Degrees Number Joint
Weight Overall Actuation Adaptive
Hand Developer of of of Coupling
(g) Size Method Grip
Joints Freedom Actuators Method
SensorHand Otto Bock 350–500 Glove sizes 2 1 1 DC Motor Fixed pinch No
(2011) [8–9] 7–8 1/4*
Vincent Hand Vincent — — 11 6 6 DC Motor- Linkage spanning Yes†
(2010) [10] Systems Worm Gear MCP to PIP
iLimb Touch 450–615 180–182 mm long, 11 6 5 DC Motor- Tendon linking Yes†
(2009) [11] Bionics 80–75 mm wide, Worm Gear MCP to PIP
35–41 mm thick
iLimb Pulse Touch 460–465 180–182 mm long, 11 6 5 DC Motor- Tendon linking Yes†
(2010) [11] Bionics 80–75 mm wide, Worm Gear MCP to PIP
35–45 mm thick
Bebionic RSL 495–539 198 mm long, 11 6 5 DC Motor- Linkage spanning Yes†
(2011) [12] Steeper 90 mm wide, Lead Screw MCP to PIP
50 mm thick
Bebionic v2 RSL 495–539 190–200 mm long, 11 6 5 DC Motor- Linkage spanning Yes†
(2011) [12] Steeper 84–92 mm wide, Lead Screw MCP to PIP
50 mm thick
Michelangelo Otto Bock ~420 — 6 2 2 — Cam design with No
(2012) [13] links to all
fingers
*Otto Bock glove sizes measured in inches from base of palm to tip of middle finger.
†Adaptive grip accomplished through electronic torque control, others from adaptive mechanical coupling.
DC = direct current, MCP = metacarpal phalange, PIP = proximal interphalange.
Table 2.
Published grip and kinematic characteristics of commercial prosthetic hands.
Grip Force Range of Motion Grasp Type
Precision Power Lateral MCP PIP DIP Thumb Thumb Thumb
Hand Finger/Grasp Achievable
Grasp Grasp Pinch Joints Joints Joints Flexion Circumduction Circumduction
Speed Grasps
(N) (N) (N) (°) (°) (°) (°) (°) Axis
SensorHand NA 100 NA 0–70* NA NA 0–70* NA None Up to Power
(2011) [8–9] 300 mm/s
at tip
Vincent Hand — — — 0–90* 0–100* NA — — Parallel with — Power, precision,
(2010) [10] wrist axis lateral, hook,
finger-point
iLimb 10.8 — 17–19.6 0–90* 0–90* ~20 0–60* 0–95* Parallel with 200 mm/s Power, precision,
(2009) [11] wrist axis lateral, hook,
finger-point
iLimb Pulse — 136 — 0–90* 0–90* ~20 0–60* 0–95* Parallel with 1.2 s (power grasp) Power, precision,
(2010) [11] wrist axis lateral, hook,
finger-point
Bebionic 34 75 15 0–90 10–90 ~20 — 0–68 Parallel with 1.9 s (power grasp), Power, precision,
(2011) [12] (tripod) wrist axis 0.8 s (tripod grasp), lateral, hook,
1.5–1.7 s (key grasp) finger-point
Bebionic v2 34 75 15 0–90* 0–90* ~20 — 0–68 Parallel with 0.9 s (power grasp), Power, precision,
(2011) [12] (tripod) wrist axis 0.4 s (tripod grasp), lateral, hook,
0.9 s (key grasp) finger-point
Michelangelo 70 NA 60 0–35* NA NA — — Compound — Opposition,
(2012) [13] axis lateral, and
neutral mode
*Estimated based on images and videos.
DIP = distal interphalange, MCP = metacarpal phalange, NA = not applicable, PIP = proximal interphalange.
602
Table 3. RESULTS
Measured commercial entire hand system weight (g).
Small Finger Large Finger Entire System
Hand Weight
Weight Weight Weight
Vincent 29–31 35–37 — The commercial hand weights are presented based on
iLimb 48 52 — the weight of the entire system required to be carried by
iLimb Pulse — — 539* the user. For the iLimb Pulse and Bebionic v2 hands, this
Bebionic v2 — — 527 includes the battery, controller, two force sensing resistors
Michelangelo — — 746* (used to simulate electromyography electrodes), and the
*Includes protective sleeve. distal side of the Otto Bock Electronic quick-disconnect
wrist unit. The Michelangelo hand entire system weight
includes the hand with protective sleeve (498 g), a much
Table 4.
Motor specifications for commercial hands. larger battery (143 g), controller (14 g), and an Axon
Gear Ratio, Rotation wrist adapter (91 g). The Vincent fingers have
Motor three different-sized distal segment attachments that allow
Hand Motor to
Type
MCP Joint the same base to be used for the three large fingers of the
Vincent Maxon 1017 — hand. The distal segment is illustrated in Figure 2(a).
iLimb Maxon RE 10 4.5 V 1.5 1600:1 Each of the end segments weighs 2 to 4 g.
W Part # 118394
iLimb Pulse Maxon RE 10 4.5 V 1.5 1600:1 Actuation Method
W Part # 118394
Bebionic Custom Linear Drive —
from Reliance Preci- Finger Kinematics
sion Mechatronics Unlike human hands, five of the six commercial
Bebionic v2 Custom Linear Drive — hands tested feature a proximal joint, similar to the
from Reliance Preci- human metacarpal phalange (MCP), and a single distal
sion Mechatronics joint that takes the form of both the human proximal
Michelangelo Custom Modified — interphalange (PIP), and distal interphalange (DIP). An
Maxon EC45 additional feature on the distal finger segment gives the
MCP = metacarpal phalange.
look of the DIP joint in the iLimb and Bebionic fingers.
The Michelangelo fingers consist of a single finger seg-
compare and discuss details regarding the hand designs. ment actuated only at a single point like that of the
Our analysis of each of these hands allows us to discuss human MCP joint and seen in Figure 2(d).
the hands side by side in a more consistent manner. Addi- Instead of actuating each joint of the fingers indepen-
tionally, experimental analysis allowed us to make obser- dently, the fingers of the iLimb, Vincent, Bebionic, and
vations regarding the kinematics of each hand that would Bebionic v2 fingers have a fixed movement relative to
have been unobtainable otherwise. each other. Figure 2 illustrates the mechanism used to
define the fixed relationship between the joint motions.
Tested Commercial Hand Systems Although these hands use a form of a four-bar linkage,
Elements of the six commercial prosthetic hands each has a distinct method of coupling the motion of the
shown in Figure 1 were acquired and tested to measure PIP to the motion of the MCP joint. The Vincent finger
their performance characteristics. The iLimb Pulse, Bebi- (Figure 2(a)) uses two externally located wire links
onic, Bebionic v2, and Michelangelo hands were tested in mounted between the finger base and the distal link. This
a fully assembled hand configuration. The iLimb Prodig- four-bar linkage mechanism, as illustrated in Figure 2(a)
its (same fingers and control system as standard iLimb) (bottom), is common among fully actuated robotic finger
and Vincent Hand finger performance characteristics were designs. The iLimb and iLimb Pulse hands use a tendon
determined through testing of a set of four fingers con- system in which a loop of fibrous cable is wrapped around
nected to a nonstandard palm mount using the same con- a bearing surface mounted to the finger base. The distal
troller and battery as the original entire hand system. end of the tendon loop is attached to the distal link and
603
Figure 3.
Vincent (Vincent Systems), iLimb (Touch Bionics), and Bebionic
v2 (RSL Steeper) hands feature linear relationship between
metacarpal phalange (MCP) and proximal interphalange (PIP)
joints during flexion/extension motion.
Figure 4.
(a) Drive mechanism of Michelangelo hand (Otto Bock). Center drive element controls flexion of all four fingers and thumb. Second
motor (which actuates against bronze worm gear) independently controls abduction/adduction of thumb. (b) Vincent finger motor (Vin-
cent Systems) is housed in proximal phalange and rotates worm against fixed worm gear to flex finger. (c) iLimb finger (Touch Bionics)
is actuated in same manner as Vincent finger but uses set of bevel gears between motor and worm drive. MCP = metacarpal phalange.
entire hand was commanded to close at full power and The grasp force was measured on the commercial
then released. Although there is a force peak at impact, the hands using pinch meters for precision grasps and a grip
constant holding force is the value presented in Tables 6 dynamometer for lateral grasp and power grasps. Each
and 7. The Vincent and iLimb Pulse hands use an addi- device was calibrated over the range of loads experienced
tional pulse mode to increase the individual finger holding
during each test. The individual finger holding force was
force. After a set period of time of motor stall, quick pulses
of power are sent to the motor. The effect is to “ratchet” the not measured for the Michelangelo hand since all digits
system to a higher capable holding force than was previ- are actuated by a central drive as opposed to a single
ously experienced. The pulse mode increased the holding drive per finger in the other commercial hands.
force of an individual finger in the Vincent hand by
an average of 69.5 percent and in the iLimb Pulse by an Compliance
average of 91.5 percent. However, the pulse mode greatly Each hand design features a mechanical element that
reduces battery life. allows for a certain level of compliance in the flexion
direction. This type of feature helps to prevent the fingers
Table 6. from breaking under any inadvertent contact, forcing the
Individual finger holding force at tip. fingers to close. The Vincent finger features a unique
Force Number of Standard bend in the links connecting the base and distal segment.
Finger
(N) Trials Deviation
Vincent Large 4.82 or 8.44* 14 or 8* 0.8 or 1.3 * The bend allows it to act like a series of elastic elements
(ring, middle, and index) and enables the distal link to move under excess force
Vincent Small (little) 3.00 2 0.1 with the MCP joint remaining fixed. The iLimb and
iLimb Large (middle) 7.66 2 0.2
iLimb Pulse hands use a simple spring and tendon drive
iLimb Med (index/ring) 5.39 4 0.1
iLimb Small (little) 5.17 2 0.1
that allows the distal link to flex inward independent of
iLimb Pulse Med (index) 4.15 or 6.54* 1 — the MCP joint. The Bebionic and Bebionic v2 are the
iLimb Pulse Large (middle) 3.09 or 6.24* 2 or 2* 0.7 or 0.4* only hands that allow for compliance in both the MCP
iLimb Pulse Med (ring) 6.43 or 11.18* 2 or 2* 0 or 0.3* and DIP joints. Although they are rigidly coupled to each
iLimb Pulse Small (little) 4.09 or 8.56* 2 or 2* 0.1 or 0*
Bebionic (index) 12.47 1 —
other, the actuator is connected to the proximal link
Bebionic (middle) 12.25 2 1.0 through a pinned slot. If the finger is forced in the flexion
Bebionic (ring) 12.53 2 1.1 direction, the pin simply rides up the slot, allowing the
Bebionic Small (little) 16.11 2 0.2 MCP and DIP joints to flex inwards. Figure 2(a) shows
Bebionic v2 Large 14.5 2 1.2 the curved linkages of the Vincent finger and the pinned
(ring, middle, and index)
*Holdingforce after pulse mode.
actuation slot of the Bebionic v2 finger. Figure 5 shows
med = medium. the direction of compliance and actuation linkage of the
Table 7.
Overall grasp holding force during grasp postures.
Lateral Grasp Palmer Grasp Power Grasp
Hand Total Force Number Standard Total Force Number Standard Total Force Number of Standard
(N) of Trials Deviation (N) of Trials Deviation (N) Trials Deviation
iLimb Pulse 17.04 or 3 or 3* 2.8 or 2.0* 10.82 or 2 or 2* 0.5 or 0.3* Large Grip: Large Grip: Large Grip: —
32.10 * 17.11 * 65.25 or 71.44* 1 or 2 or 4.0*
Small Grip: Small Grip: 1* Small Grip: —*
50.81*
Bebionic 17.61 1 — 29.47 1 — 77.37 1 —
Bebionic v2 16.4 4 3.2 22.53 4 1.5 62.4 6 10.3
Michelangelo 50.84 4 3.1 78.14 8 4.4 Grasp Type Grasp Type Grasp Type
Unachievable Unachievable Unachievable
*Holding force after pulse mode.
— indicates no standard deviation because only one trial performed.
606
DISCUSSION
Figure 6.
Illustration of circumduction axis location for Bebionic v2 thumb (RSL Steeper) (shown from bottom view).
Tables 1 and 8 show the weight of both current pros- hands, shown in Table 1, are designed to be covered with
thetic hands and research hands designed for use in pros- a silicone glove to enhance the cosmetic appearance of
theses. A range of 350 to 615 g is seen in current the prosthesis. Since prosthetic hands are sized according
commercial prosthetics and 350 to 2,200 g in research- to human hand measurements (and commonly based on
based hands. Data presented in the tables are based on direct measurements of the patient’s able hand), the pros-
values published by the various research groups and do thetic hand structure, including cosmetic covering, should
not reflect a consistent comparison of weight. For some have a length between 180 and 198 mm and a width of 75
hands, the entire actuation and control system including to 90 mm to match normal human hand size [11].
batteries and wrist attachment is included in the total
weight of the hand. Others only consider the weight of Actuation Properties
the hand itself and not the external computing or power
sources for operation. Finger kinematics. Anatomically correct finger kine-
matics are a goal in mechanical design of prosthetic hands.
Within the prosthetics community, no set specifica-
However, there is a trade-off between anatomical correct-
tion exists for the maximum weight of the prosthesis. The
ness and robustness, weight, complexity, and cost. In many
only agreed upon specification is to minimize weight in
of the hands reviewed in this article, there are more joints
general. Ultimately, the weight will relate to the required
than number of actuators. Often, numerous joints will be
size and capabilities of the hand. According to Pons et
coupled to act as a single compound motion where only the
al., an adult-sized prosthetic hand should weigh less than
actuator position, for example, must be known to determine
400 g [20]. Kay and Rakic have set a requirement that the
the position of all joints that are coupled together. A distinct
entire hand including cosmetic glove should remain
set of movements that can be described by a single parame-
under 370 g [21], while other groups, including Light and
ter is considered a single DOF. The four fingers of the
Chappell [22] and Vinet et al. [23], believe a 500 g
MANUS-Hand (collaboration between Consejo Superior
weight limit is appropriate.
de Investigaciones Científicas, Argana del Rey, Spain;
Hand size. For an anthropomorphic prosthesis, it is Ketholiek Universiteit Leuven, Belgium; Centro de Recu-
natural for the envelope of the hand to replicate the size peracion de Minusvalidos Fisicos, IMSERSO, Spain; Alor-
and shape that is natural to the user. All of the myoelectric man Advanced Medical Technologies Ltd, Israel; and
608
Table 8.
Published general characteristics of 13 research hands with applications in prosthetics.
Number Degrees Number
Weight Overall Actuation Joint Coupling Adaptive
Hand Developer of of of
(g) Size Method Method Grip
Joints Freedom Actuators
TBM Hand* University of 280 146 mm long, 15 6 1 DC Motor with Compliant Yes
(1999) [24] Toronto 65 mm wide, Linear Ball springs
25 mm thick Screw
Remedi Hand University of 400 Similar to human 14 6 6 DC Motor Coupled MCP, No
(2000) [22] Southampton hand (Maxon) DIP, PIP
RTR II ARTS/Mitech 350 — 9 9 2 DC Motors Tendon and Yes
(2002) [25] Laboratories free-spinning
(Pisa Italy) pulleys
MANUS-Hand Spain/Belgium/ 1200 — 9 3 2 Brushless DC Fixed coupling No†
(2004) [20] Israel Motors of MCP, PIP,
and DIP
DLR/HIT I DLR German Space 2,200 1.5× human hand 17 13 13 Brushless DC 1:1 coupling of No
(2004) [26] Agency, Harbin Motors with two distal flexion
Institute of Planetary Drive joints
Technology
DLR/HIT II DLR German 1,500 Human hand size 20 15 15 Brushless DC 1:1 coupling of No
(2008) [26–27] Space Agency Motors with two distal flexion
Harmonic joints
Drive
UB Hand 3 University of — Human hand size 18 15 16 HiTec Servos PIP and DIP No
(2005) [28] Bologna, Italy coupled in ring,
little, and thumb
UNB Hand University of New — Size 7.5 10 5 3† DC Motors Fixed coupling of Yes
(2010) [29–30] Brunswick (MicroMo PIP to MCP
1724)
FluidHand III Forschungszen- 400 Similar to human 8 8 1 pump, Pressurized Distributed Yes
(2009) [31] trum Karlsuhe hand 5 valves fluid pressure
GmbH (KIT)
Smarthand ARTS Laboratory, 520 12 mm longer and 16 16 4 DC Motors Tendon/spring Yes
(2009) [2,32] Pontedera Italy 8 mm thicker (Faulhaber) based
than 50% male
Keio Hand Keio University, 730 320 mm length 15 15 1 Ultrasonic Motor Single tendon Yes
(2008) [33] Yokohama, Japan (with motor), for each finger
120 mm fingers
Vanderbilt Hand Vanderbilt 580 190 mm long, 16 16 5 Brushed DC Single cable for Yes
(2009) [34] University 330 mm with Servomotors each finger
motors, 75 mm mounted in
wide Forearm
LO/SH University of — — 8 4 2 DC Motors Wiffle tree along Yes
Southampton Southampton finger
Hand (2001) [35]
*Designed for children.
†Two degrees of freedom
of thumb controlled through single motor.
DC = direct current, DIP = distal interphalange, MCP = metacarpal phalange, PIP = proximal interphalange.
Advanced Material Technologies N.V., Belgium) [20] are tion of the joints since they are dependent on the contact
considered one DOF (despite having 8 joints) since they are state of each finger link with the object. These mechanisms
directly coupled to one another. This is an example of a rig- are considered adaptive because, when they are used in a
idly coupled hand. Another way of coupling is through hand, they allow multiple links of the fingers to passively
adaptive underactuation, in which a single actuator controls adapt to the shape and location of an object with a single
a number of independent DOFs [36]. In this sense, the sin- actuator [37–38]. Examples of adaptive finger designs in
gle actuator parameter cannot be used to describe the posi- prosthetics include a single tendon routed across multiple
609
joints, such as in the Vanderbilt hand (Center for Intelligent capabilities but require more complex motor control
Mechatronics, Vanderbilt University; Nashville, Tennessee) schemes. Brushless motors typically include sensors that
[34] and RTR-II (ARTS/Mitech Labs, Scuola Superiore can provide additional position feedback. Moreover, as
Sant’Anna; Pisa, Italy, and Centro INAIL RTR; Viareggio, control electronics continue to shrink in size, brushless
Italy) [25], or the compliant spring connections used in the DC motors will likely become the dominant motor choice.
TBM hand (Department of Mechanical and Industrial Engi- All DC motors naturally produce excessive speed and
neering and Institute of Biometerials and Biomedical Engi- insufficient torque for use in prosthetic devices. There-
neering, University of Toronto; Toronto, Canada, and fore, drive reductions are necessary to reduce the speed
Rehabilitation Engineering Department, Bloorview Mac- and increase the torque provided by the actuator [1].
Millian Center; Toronto, Canada) [24] and Smarthand In order to reduce the speed and increase the limited
(ARTS Laboratory, Sculuola Superiore Sant’Anna; Pont- torque from these motors, gearing, lead screws, and even
edera, Italy) [2,32]. harmonic drives may be used. The iLimb and Vincent
Tables 2 and 9 show the range of finger motion for hands package a single motor and gear train in the proximal
both commercial and research hands. For commercial phalange of each finger. The FluidHand III (Forschung-
hands, the PIP and MCP joints exhibit similar ranges of szentrum Karlsruhe GmbH; Eggenstein-Leopoldshafen,
motion to the human hand. The DIP joint, however, is Germany) uses a small DC motor to drive a small hydraulic
usually fixed at 20°. pump housed within the palm of the hand [29]. Five inde-
pendent valves then transmit pressure to bellows located
Thumb kinematics. The thumb design in an anthro- at each joint. The advantage of using a pressure-based sys-
pomorphic prosthetic hand is critical since the thumb tem is the compliance associated with each finger joint,
accounts for arguably 40 percent of the entire functional- which allows the system to survive sudden impacts. Many
ity of the human hand [39]. In most of the prosthetic of the hands incorporate nonbackdriveable mechanisms
hands described in this article, the thumb is actuated in (NBDMs) between the motor and the flexion of the fingers.
flexion/extension (simple closing or opening) and along NBDMs allow the finger to maintain high grip forces
the circumduction axis. The circumduction rotation of the (assisted by compliance in the mechanism) without contin-
thumb is the movement required to alternate between a ued current draw from the battery. The most common
lateral grasp and a power or precision grasp. An analysis NBDMs include lead screws, worm drives, and roller
of human hand kinematics shows an average circumduc- clutches. See Weir [1] and Controzzi et al. [41] for addi-
tion motion of 90.2°, which is achieved through a combi- tional information regarding NBDMs.
nation of three joints at the base of the thumb [40]. As
can be seen in Tables 2 and 9, the circumduction axis of Grip force. Most activities of daily living (ADLs)
current hands is not always oriented parallel with the require fast speed and low grip force (e.g., typing, gestur-
wrist rotation axis. By angling this axis ventrally or dor- ing). However, tasks that require low speeds and high
sally, thumb flexion and circumduction rotation can be grip force occur often enough that a prosthetic hand must
jointly approximated in a single DOF. This can be benefi- enable the user to perform such tasks (e.g., opening door
cial to achieve desired hand openings and a more anthro- with handle, unscrewing jar lid).
pomorphic motion for precision, power, and lateral grasp The grip force able to be exerted by a hand on an
patterns while keeping complexity low. The coupling can object is largely a function of the hand posture, object
also help the timing of the grasp if all of the fingers are geometry, and transmission method. In particular, pros-
actuated simultaneously. Further discussion of the role of thetic hands like the Hosmer Hook (Hosmer; Campbell,
the thumb circumduction axis can be found in other California), SensorHand [8–9], and TBM Hand [24] will
reviews [1,21,23,40]. exhibit different grasp forces depending on the size of the
object. The necessary grasp force to maintain an object
Type of actuator and drive mechanism. The most within a particular grasp is also difficult to predict
common actuator used in prosthetics today, excluding a because it is largely dependent on the friction between
body-powered harness, is a direct current (DC) motor. the fingers of the hand and the object, the number of con-
These motors are small and lightweight and can be pack- tact points, the relative locations of contact, and the
aged in the hand. Brushed DC motors are more commonly object geometry and mass properties. In a precision
used in prosthetic hands because of their ease of control. grasp, the human hand can exert an average of 95.6 N of
Brushless DC motors provide higher torque-to-weight force [1]. In power grasps, the forces can reach up to
610
Table 9.
Published grip and kinematic characteristics of 13 research hands.
Grip Force Range of Motion Grasp Type
Vanderbilt Hand 20.0 80 0–90 0–90 0–90 10 to 80 15° toward little — 225°/s, 0.4 s Power, precision,
(2009) [34] finger from wrist to close lateral, hook,
axis* finger-point
LO/SH 45.0 — — — — — — PIP joint Full close Precision/tripod
Southampton Hand only <1.2 s
(2001) [35]
*Estimated based on images and videos.
†Abduction/adduction of thumb but not rotationabout circumduction axis.
‡No independent control of fingers 3–5.
DIP = distal interphalange, MCP = metacarpal phalange, PIP = proximal interphalange.
400 N [1]. According to Heckathorne [42], a grip force of suggest a minimum grip force of 45 N for prosthetic
only 68 N is required to carry out ADLs [42]. Vinet et al. hands for practical use [23].
611
Tables 2 and 9 show the published grasp force mea- these tables are based on published information, there are
surements in three grasp configurations for common numerous ways the speeds have been described. What is
prosthetic and research hands. The more dexterous robot of most concern to the end user, however, is the amount of
hands, such as the DLR/HIT II (Institute of Robotics and time it takes to acquire an object in different possible
Mechatronics German Aerospace Center DLR; Wessling, grasp configurations. Some groups, therefore, present
Germany, and Robot Research Institute of Harbin Insti- grasp speed as a measure of time to open or close the
tute of Technology; Harbin, China) [26–27] and the UB hand. Presenting hand speed data in terms of total time to
Hand 3 (University of Bologna; Bologna, Italy) [28] have acquire an object is problematic since the metric is depen-
a lower grip force than the simpler SensorHand [8–9] and dent on the size and shape of the object. The finger flexion
MANUS-hand [20]. This trend is due to a mechanical speeds for the hands surveyed in this article ranged from
design trade-off between complexity and strength when 20 °/s (TBM hand, 4–5 s to close grasp) to 225 °/s
constrained by size. Figure 7 shows the relationship (Vanderbilt hand, 0.4 s to close). Tözeren suggests that a
between the number of actuators and the published grip 0.8 s closing time is sufficient for prosthetic hands [17].
force during a palmar/precision grasp for multiple pros- Dechev et al. states a slower 1.0 to 1.5 s closing time is
thetic hands and research hands. These results vary adequate for conducting ADLs [24]. In fact, closing
widely because of differences in actuator size, transmis- speeds that are too fast can be a substantial negative
sion ratio, and mechanism friction. because many myoelectric control schemes rely on the
user to stop the hand at the right closing position while it’s
Grasp speed. According to Pylatiuk et al., 100 per-
moving (i.e., no direct position control); excessive closing
cent of females, 76 percent of males, and 50 percent of
speed makes that substantially more difficult.
children surveyed (4, 26, and 7 subjects, respectively)
would describe the speed of their myoelectric prosthesis Achievable grasps. The typical ADLs conducted by
to be “too slow” [18]. Although the human hand can an amputee can be accomplished using a finite set of pre-
exhibit finger flexion speeds of 2,290 °/s, the typical defined grasps. These grasp patterns include power (used
speeds for everyday pick and place tasks is 172 to 200 °/s in 35% of ADLs), precision (used in 30% of ADLs), lat-
[1,42]. Tables 2 and 9 show the published grasp speeds of eral (used in 20% of ADLs), hook, tripod, and finger
numerous prosthetic hands. Since the data compiled in point [2]. Some researchers consider certain gesturing to
be important (e.g., finger counting) [2]. The full range of
distinct grasp types for the able hand is greater than 30,
and detailed descriptions of these can be found in Cut-
kosky [43]. Tables 2 and 9 show the ability of each hand
studied within this article to form these grasp patterns
without considering contact forces with the object. In
order for a hand to accomplish all seven grasping patterns
(six standard grasps plus finger counting), each individ-
ual finger flexion motion must be controlled with an
independent actuator that is not coupled to the other fin-
gers. However, removing the requirement for finger
counting can reduce this to a smaller number, particularly
if external interaction is permitted, such as a common
feature for thumb circumduction axis to be passive and
changed by the user, as is the case with the TBM, iLimb,
and Bebionic hands. A passive thumb mechanism
requires an external force to maneuver the thumb into
distinct postures and cannot be moved by the device.
Figure 7. Many hands, such as the TBM hand, attempt to accom-
Comparison between number of actuators and palmer/preci- plish as many patterns as possible with fewer than five
sion grip force in prosthetic and research hands. individual actuators. This hand uses a single actuator
612
Durability/Cycles of Use
The average myoelectric prosthetic hand user will
wear his or her device in excess of 8 h per day [18]. Any
device must therefore be robust enough to withstand pro-
longed use and comfortable enough for the user to wear
for this amount of time. The mechanical design of the
hand must consider the trade-off between durability and
robustness with weight, expense, and size. Compliant
components like conforming fingertip/palmar pads, com-
pliant actuators designs, collapsible linkage systems, etc.,
add robustness and function while not adding overdue
complexity. In general, increasing robustness typically
reduces complexity (i.e., number of DOFs, number of Figure 8.
Relationship between weight of surveyed hands and number of
components) and increases size and weight.
actuators.
According to Zheng et al., between 2,500 and 3,000
grasping motions of the unaffected dominant hand may
be performed over an 8 h period during the work day
[44]. A prosthetic device will typically undergo 120 greatly increase the weight of the hand because of the
grasping motions per day [5]. The predicted grasps of the increase in mechanical complexity to implement the cou-
prosthesis is lower than the able hand since a reduction in pling, as illustrated with the Keio (System Design and
functionality will likely result in less frequent use. A Management and Integrated Design Engineering, Keio
report by Vinet et al., intended to put forth specifications University; Yokohama, Japan) [33] and MANUS-Hand
for electromechanical hands, claimed that prostheses [16]. Figure 9 shows that the total number of joints in the
should withstand 300,000 grasping cycles and maintain hand is strongly correlated with the weight of the hand,
all of its original functionality [23]. Given the daily regardless of coupling methods. Figure 10 shows the
expected number of cycles described previously, this relationship between the total number of joints and the
would put the lifetime of the device at about 6 yr. It is number of actuators for the hands presented in this arti-
suggested that current devices last in excess of 500,000 cle. Hands lying on the dashed line have a single motor
grasp cycles between routine servicing. for each joint of the hand with no coupling between
joints. The hands that fall to the right of the dashed line
Discussion of Hand Design Trade-offs indicate that they contain some form of coupling between
Because of the strict limitations on size and weight joints. A large group of research hands are contained in
for practical prosthetic hands, the performance trade-offs the range of 15 to 20 joints, which approaches the num-
between various design options must be addressed by the ber of joints in the human hand (~30).
designer. The information presented in this article may be The choice of the number of actuators in the hand has
used to benchmark performance and compare various traditionally been based on the type of tasks that hand is
prosthetic and research hand designs. designed to achieve. In theory, a grasping hand can be
designed with 2 DOFs (actuators) since grasping is a low
Number of Actuators Versus Hand Complexity dimensional task [45] requiring a minimum of 1 to
Based on the data presented in Tables 1 and 2, a 2 DOFs to execute all functional grasps (lateral, palmar,
comparison can be made between the weight of each power). A more dexterous hand capable of a high number
hand and the number of actuators used. Figure 8 shows of grasp configurations and individual finger motions typi-
that although there may be an increase in weight of the cally requires a high number of actuators regardless of the
hand associated with the number of actuators, the cou- type of coupling used in the hand. However, clever use of
pling of multiple joints to one or two motors can still coupling strategies like whiffle tree designs or differential
613
are used more frequently. The fingers are typically cou- plishing typical tasks, high durability, light weight, and
pled to a single actuator in either a direct rigid coupling low cost. If we simply directly compare hand features
or an adaptive underactuated method. and specifications, the split hook would never be consid-
Underactuated hands present several design advan- ered the best device, but from a practical sense, it has
tages compared with rigidly linked hands. Adaptive fin- proven to remain the most common choice. Therefore,
gers can interact with objects over more locations and although comparison of hand features and specification
thereby distribute the grasping force over more contacts. can tell us about the similarities and differences in hand
Also, the mechanism can take a greater range of configu- designs, it may not tell us the full story about the level of
rations for the same number of actuators. Kargov et al. potential benefit to end users.
concluded that, while the contact forces are higher when
using a fixed coupling fingered prosthesis, the joint Design Iterations and Continual Improvement
torques of adaptive fingered hands are comparable with The design of commercial and robotic hands is a con-
the joint torques of human hands [48]. tinual process that requires numerous versions and itera-
Although some commonly used prosthetic hands tions to perfect. The hands featured in this article
allow for adaptability in grasping, a study by Bergman et represent the versions or iterations of the design that were
al. in 1992 claimed that a conventional nonadaptive pros- available to the authors at the time of publication. Multi-
thesis showed “significantly better results” regarding ple publications are made regarding the development and
width of grip, force of grip, and scores in a standardized testing of robotic hands that often contained improved
grip function test when compared with a similar adaptive performance measurements with later versions. Please
prosthesis [49]. refer to the cited documents for details regarding the
exact version of the hand specifications presented in this
Research Versus Commercial Prostheses article.
Research hands are typically “one-off” prototypes
developed as “proof of concept” devices related to a
novel design approach or as a platform to study a pros- CONCLUSIONS
thetic control method. Often times, the researchers that
produce them are not focused on or not interested in The information presented in this article serves as a
many of the details related to whether they might be com- compilation of empirical and published hand characteris-
mercially viable. Therefore, they are often not designed tics and performance measures. Within this article, we
for many of the issues relevant to successful commercial focused on the mechanical characteristics of hands with-
products, including robustness, weight, ease of mass pro- out treatment of sensing, controls, electronics, and power
duction, and cost. Along these lines, it is clear through requirements and techniques.
the successes/failures of various commercial devices and Since a hand, like any other tool, has many uses, suffi-
durability history of research hands that there are gener- cient performance for one application might not be appro-
ally design trade-offs between the complexity of the hand priate for another. It is therefore difficult to establish exact
(often including the number of DOFs) and the durability/ mechanical and performance requirements. Ultimately,
robustness of the hand. Accordingly, the least robust the selection of hand characteristics and specification is a
hands are commonly the highly dexterous and compli- choice between trade-offs in complexity, dexterity (e.g.,
cated hands, typically developed for research purposes. achievable grasps), weight, and control methods. Further-
more, each of these measures are subject to the patients’
Features and Specifications Versus Practical Functionality exact needs, including the nature and level of their ampu-
Although numerous prosthetic hands can be com- tation, as well as level of activity, professional needs, and
pared using their relevant features and specifications, the others. The entire prosthesis must work as a system to
actual goal is to create a practical device for users. Even facilitate the accomplishment of tasks.
with the technology available today, the most frequently A set of clinical standards for performance, including
used prosthetic terminal devices are still the split hook techniques for evaluating anthropomorphic hand designs,
type devices (such as the Hosmer Hook). The reasons for would be beneficial and is planned for future work by the
this include its practicality and ease of use for accom- authors. It is clear from this review that the current
615
9. SensorHand speed [Internet]. Duderstadt (Germany): Otto 24. Dechev N, Cleghorn WL, Naumann S. Multiple finger,
Bock; 2013. Available from: passive adaptive grasp prosthetic hand. Mechanism Mach
http://www.ottobock.com/cps/rde/xchg/ob_com_en/hs.xsl/ Theory. 2001;36(10):1157–73.
3652.html http://dx.doi.org/10.1016/S0094-114X(01)00035-0
10. VINCENT hand [Internet]. Weingarten (Germany): Vin- 25. Massa B, Roccella S, Carrozza MC, Dario P. Design and
cent Systems; 2013. Available from: development of an underactuated prosthetic hand. Proceed-
http://handprothese.de/vincent-hand/ ings of the 2002 IEEE International Conference on Robot-
11. Touch Bionics web site [Internet]. Mansfield (MA): Touch ics and Automation; 2002 May 11–15; Washington, DC.
Bionics Inc; 2013. Available from: p. 3374–79.
http://www.touchbionics.com/ 26. Liu H, Wu K, Meusel P, Seitz N, Hirzinger G, Jin MH, Liu
12. RSL Steeper web site [Internet]. Leeds (United Kingdom): YW, Fan SW, Lan T, Chen ZP. Multisensory five-fingered
RSL Steeper; 2013. Available from: http://rslsteeper.com/ dexterous hand: The DLR/HIT Hand II. Proceedings of the
13. Michelangelo operation manual. Duderstadt (Germany): 2008 IEEE/RSJ International Conference on Intelligent
Otto Bock; 2012. Robots and System; 2008 Sep 22–26; Nice, France.
14. Ingram JN, Körding KP, Howard IS, Wolpert DM. The sta- p. 3692–97.
tistics of natural hand movements. Exp Brain Res. 2008; 27. DRL Institute of Robotics and Mechatronics. Datasheet of
188(2):223–36. [PMID:18369608] DLR Hand II [Internet]. Cologne (Germany): DLR; 2011.
http://dx.doi.org/10.1007/s00221-008-1355-3 Available from:
15. Maxon Motor. Motor data [Internet]. Sachseln (Switzer- http://www.dlr.de/rm/en/desktopdefault.aspx/tabid-3802/
land): Maxon Motor; 2013. Available from: 6102_read-8922/
http://shop.maxonmotor.com/ishop/app 28. Lotti F, Tiezzi P, Vassura G, Biagiotti L, Palli G, Melchiorri
16. Chandler RF, Clauser DE, McMconville JT, Reynolds HM, C. Development of UB Hand 3: Early results. Proceedings
Young JW. Investigation of inertial properties of the human of the 2005 IEEE International Conference on Robotics
hand. Washington (DC): U.S. Department of Transporta- and Automation; 2005 Apr; Barcelona, Spain. p. 4488–93.
tion; 1975 Mar. Report No. DOT HS-801 430. 29. Losier Y, Clawson A, Wilson A, Scheme E, Englehard K,
17. Tözeren A. Human body dynamics: Classical mechanic Kyberg P, Hudgins B. An overview of the UNB hand sys-
and human movement. New York (NY): Springer; 2000. tem. Proceeding of the 2011 Myoelectric Controls/Powered
18. Pylatiuk C, Schulz S, Doderlein L. Results of an Internet Prosthetics Symposium; 2011; New Brunswick, Canada.
survey of myoelectric prosthetic hand users. Prosthet 30. Clawson A, Segil J, Jones B, Losier Y, Kyberg PJ, Weir R.
Orthot Int. 2007;31(4):362–70. [PMID:18050007] Mechanical design of a multifunction hand prosthesis sys-
http://dx.doi.org/10.1080/03093640601061265 tem—The UNB Hand. Proceedings of the 13th ISPO
19. Jönsson S, Caine-Winterberger K, Brånemark R. Osseoin- World Congress; 2010 May 10–15; Leipzig, Germany.
tegration amputation prostheses on the upper limbs: meth- 31. Gaiser IN, Pylatiuk C, Schulz S, Kargov A, Oberle R, Wer-
ods, prosthetics and rehabilitation. Prosthet Orthot Int. ner T. The FLUIDHAND III: A multifunctional prosthetic
2011;35(2):190–200. [PMID:21697201] hand. Am Acad Orthot Prosthet. 2009;21(2):91–96.
http://dx.doi.org/10.1177/0309364611409003 32. Cipriani C, Controzzi M, Carrozza MC. Mechanical design
20. Pons JL, Rocon E, Ceres R, Reynaerts D, Saro B, Levin S, of a transradial cybernetic hand. Proceedings of the 2008
Van Moorleghem W. The MANUS-HAND* Dextrous IEEE/RSJ International Conference on Intelligent Robots
Robotics Upper Limb Prothesis: Mechanical and manipula- and Systems; 2008 Sep 22–26; Nice, France. p. 576–81.
tion aspects. Auton Robots. 2004;16:143–63. 33. Kamikawa Y, Maeno T. Underactuated five-finger pros-
http://dx.doi.org/10.1023/B:AURO.0000016862.38337.f1 thetic hand inspired by grasping force distribution of
21. Kay HW, Rakic M. Specifications for electromechanical humans. Proceedings of the 2008 IEEE/RSJ International
hands. Proceedings of the 4th International Symposium on Conference on Intelligent Robots and Systems; 2008 Sep
the External Control of Human Extremities; 1972 Aug 28– 22–26; Nice, France. p. 717–22.
Sep 2; Belgrade, Yugoslavia. p. 137–55. 34. Dalley SA, Wiste TE, Withrow TJ, Goldfarb M. Design of
22. Light CM, Chappell PH. Development of a lightweight and a multifunctional anthropomorphic prosthetic hand with
adaptable multiple-axis hand prosthesis. Med Eng Phys. extrinsic actuation. IEEE/ASME Transactions on Mecha-
2000;22(10):679–84. [PMID:11334753] tronics. 2009;14(6):699–706.
http://dx.doi.org/10.1016/S1350-4533(01)00017-0 http://dx.doi.org/10.1109/TMECH.2009.2033113
23. Vinet R, Lozac’h Y, Beaudry N, Drouin G. Design method- 35. Kyberd PJ, Light C, Chappell PH, Nightingale JM, What-
ology for a multifunctional hand prosthesis. J Rehabil Res ley D, Evans M. The design of anthropomorphic prosthetic
Dev. 1995;32(4):316–24. [PMID:8770796] hands: A study of the Southampton Hand. Robotica. 2001;
617